Enable sentences with features about nmea

[PegasisForever]

I guess this will be a breaking change? If we make all-sentences the default feature, applications use default-features = false for no-std support will break

[elpiel]

Valid point. Yes it will be, but we'll bump version to 0.5.
We can also include Changelog but I'm not afraid to break the build from minor version updates.

default = ["std", "all-sentences"]
all-sentences = ["...", "vendor-specific"]
vendor-specific = [".."]
...
..
.

[Dushistov]

What about usage of macros instead of features?

Something like

generate_parser![SentenceType::AAM, SentenceType::ABK];

and that generate nmea parser that parse only AAM and ABK messages,
so compiler+linker in release mode can remove unused part of nmea crate?

Plus it would be nice implement the similar fnctionality for Nmea::create_for_navigation.

This variant of dealing with redundant code allows the compiler to solve the unnecessary code problem,
instead of manually marking of each sentence with feature.

[elpiel]

@Dushistov macros might make it more difficult to use the crate and contribute to the project IMO.
Also, not all tools/ides/editors play nicely with macros.
On no_std crates, I've seen macros used to implement different peripherals but not exposing the internal macros for users to define their own.

It's tedious process to add these features now but explicitly listing them sounds like a very good approach.
Due to the way parsers work, I suppose we can even add "Feature not enabled" error when a sentences is passed which can be parsed but the feature hasn't been enabled.

[elpiel]

There's one more additional benefit that I though of.
We can leave the parsers themselves and just exclude the calling of the parser in the parse_str function.

Compiler will still optimize and remove the unused parser/code, however, if you want to parse a specific sentences in your project which you have excluded from parse_str, then it's your choice and you can do it without enabling the feature.

[ekuinox]

In my case, I use my own ParseResult-like enumeration type combined with some public parsers instead of using ParseResult. So I think it is a good change to be able to exclude unused sentences from the compilation by means of a feature flag.